U.S. patent number 8,699,635 [Application Number 12/617,537] was granted by the patent office on 2014-04-15 for frame boundary detection.
This patent grant is currently assigned to Cambridge Silicon Radio Limited. The grantee listed for this patent is Andrei Popescu, Chunyang Yu. Invention is credited to Andrei Popescu, Chunyang Yu.
United States Patent |
8,699,635 |
Yu , et al. |
April 15, 2014 |
**Please see images for:
( Certificate of Correction ) ** |
Frame boundary detection
Abstract
A method of WLAN frame detection in a received signal, wherein
the frame comprises first and second training sequences and the
method comprises auto-correlating the signal with a delayed version
of itself to establish a first frame boundary estimate based on
behavior of the autocorrelation result due to the inclusion of the
first training sequence in the frame, cross-correlating the signal
with a copy of the second training sequence at a range of time
offsets in order to generate a first cross-correlation profile,
classifying the first cross-correlation profile into one of a
number of categories, establishing a second frame boundary estimate
from the first cross-correlation profile in a manner dependent upon
the category assigned to the first cross-correlation profile and
determining a refined frame boundary estimate on the basis of a
consideration of the first and second frame boundary estimates.
Apparatus for performing the method is also described.
Inventors: |
Yu; Chunyang (Cambridge,
GB), Popescu; Andrei (Cambridge, GB) |
Applicant: |
Name |
City |
State |
Country |
Type |
Yu; Chunyang
Popescu; Andrei |
Cambridge
Cambridge |
N/A
N/A |
GB
GB |
|
|
Assignee: |
Cambridge Silicon Radio Limited
(Cambridge, GB)
|
Family
ID: |
43974161 |
Appl.
No.: |
12/617,537 |
Filed: |
November 12, 2009 |
Prior Publication Data
|
|
|
|
Document
Identifier |
Publication Date |
|
US 20110110470 A1 |
May 12, 2011 |
|
Current U.S.
Class: |
375/343; 370/210;
375/260; 375/340; 370/310; 370/208; 375/355 |
Current CPC
Class: |
H04L
27/2675 (20130101); H04L 27/2656 (20130101); H04L
27/2657 (20130101); H04L 27/2662 (20130101); H04L
5/0023 (20130101) |
Current International
Class: |
H03D
1/00 (20060101) |
Field of
Search: |
;375/343,340,355,260 |
References Cited
[Referenced By]
U.S. Patent Documents
Primary Examiner: Puente; Eva
Attorney, Agent or Firm: RatnerPrestia
Claims
The invention claimed is:
1. A method of WLAN frame detection in a received signal, wherein
the frame comprises first and second training sequences and the
method comprises auto-correlating the signal with a delayed version
of itself to establish a first frame boundary estimate based on
behaviour of the autocorrelation result due to the inclusion of the
first training sequence in the frame, cross-correlating the signal
with a copy of the second training sequence at a range of time
offsets in order to generate a first cross-correlation profile,
classifying the first cross-correlation profile into one of at
least three categories, establishing a second frame boundary
estimate from the first cross-correlation profile in a manner
dependent upon a category of the at least three categories into
which the first cross-correlation profile is classified and
determining a refined frame boundary estimate on the basis of the
first and second frame boundary estimates.
2. A method according to claim 1, wherein the frame comprises a
third training sequence and the method further comprises
cross-correlating the signal with a copy of the third training
sequence at a range of time delays in order to generate a second
cross-correlation profile and boosting the first cross-correlation
profile prior to its classification by coherently adding the second
cross-correlation profile into the first cross-correlation
profile.
3. A method according to claim 1, wherein determining the refined
frame boundary estimate comprises selecting, as the refined frame
boundary estimate, the first frame boundary estimate, the second
frame boundary estimate or a weighted combination of the first and
second frame boundary estimates.
4. A method according to claim 3, wherein the selection of the
refined frame boundary estimate depends on a peak magnitude in the
auto-correlation result.
5. A method according to claim 3, wherein the selection of the
refined frame boundary estimate depends on a difference between the
first and second frame boundary estimates.
6. A method according to claim 3, wherein the selection of the
refined frame boundary estimate depends on a peak magnitude in the
first cross-correlation profile.
7. A method according to claim 1, wherein one of the at least three
categories is where the first cross-correlation profile has a
single peak and for that category the second frame boundary
estimate is established as the position of that peak.
8. A method according to claim 1, wherein one of the at least three
categories is where the first cross-correlation profile has just
two peaks and for that category the second frame boundary estimate
is established as the position of the later of those two peaks.
9. A method according to claim 8, wherein the first
cross-correlation profile is smoothed before establishing the
second frame boundary estimate.
10. A method according to claim 1, wherein one of the at least
three categories is where the first cross-correlation profile has
more than two peaks and for that category the second frame boundary
estimate is established in dependence upon the positions of the
three largest peaks.
11. A method according to claim 1, wherein one of the at least
three categories is a residual category for the case where the
first cross-correlation profile fits no other category and for the
residual category the second frame boundary estimate is established
as the position of the maximum in the profile after smoothing.
12. A method of WLAN frame detection in a received signal, wherein
the frame comprises first and second training sequences and the
method comprises auto-correlating the signal with a delayed version
of itself to establish a first frame boundary estimate based on
behaviour of the autocorrelation result due to the inclusion of the
first training sequence in the frame, cross-correlating the signal
with a copy of the second training sequence at a range of time
offsets in order to generate a first cross-correlation profile,
classifying the first cross-correlation profile into one of a
number of categories based on a number of peaks in the first
cross-correlation profile, establishing a second frame boundary
estimate from the first cross-correlation profile in a manner
dependent upon the category into which the first cross-correlation
profile is classified and determining a refined frame boundary
estimate on the basis of the first and second frame boundary
estimates.
13. A method according to claim 12, wherein the frame comprises a
third training sequence and the method further comprises
cross-correlating the signal with a copy of the third training
sequence at a range of time delays in order to generate a second
cross-correlation profile and boosting the first cross-correlation
profile prior to its classification by coherently adding the second
cross-correlation profile into the first cross-correlation
profile.
14. A method according to claim 12, wherein determining the refined
frame boundary estimate comprises selecting, as the refined frame
boundary estimate, the first frame boundary estimate, the second
frame boundary estimate or a weighted combination of the first and
second frame boundary estimates.
15. A method according to claim 14, wherein the selection of the
refined frame boundary estimate depends on a peak magnitude in at
least one of the auto-correlation result and the first
cross-correlation profile.
16. A method according to claim 14, wherein the selection of the
refined frame boundary estimate depends on a difference between the
first and second frame boundary estimates.
17. A method according to claim 12, wherein one of the categories
is where the first cross-correlation profile has a single peak and
for that category the second frame boundary estimate is established
as the position of that peak.
18. A method according to claim 12, wherein one of the categories
is where the first cross-correlation profile has just two peaks and
for that category the second frame boundary estimate is established
as the position of the later of those two peaks.
19. A method according to claim 12, wherein one of the categories
is where the first cross-correlation profile has more than two
peaks and for that category the second frame boundary estimate is
established in dependence upon the positions of the three largest
peaks.
20. A method according to claim 12, wherein one of the categories
is a residual category for the case where the first
cross-correlation profile fits no other category and for the
residual category the second frame boundary estimate is established
as the position of the maximum in the profile after smoothing.
Description
FIELD OF THE INVENTION
The invention relates to the field of communications, and in
particular to the deduction of frame boundaries within a received
signal.
BACKGROUND OF THE INVENTION
According to the IEEE 802.11 standard and in particular its
802.11n-2009 amendment, a physical layer OFDM frame (a `WLAN frame`
or a `frame`) begins with an STF (short training field) followed by
an LTF (long training field). In frames with so-called `legacy`
format and frames with mixed mode format, it is L-STF and L-LTF; in
frames with green field format, it is HT-GF-STF and HT-GF-LTF1.
An important role of the STF and first LTF is to allow a receiver
to estimate the start time (to establish time synchronization) for
a received frame. Time synchronization is a sensitive part of
802.11 OFDM receiver design; large synchronisation timing errors
introduce inter-symbol interference (ISI) that severely degrades
reception. However, because OFDM symbols include a guard interval,
small synchronization timing errors may be tolerated without much
degradation of the receiver performance.
The STF consists of 10 identical short training symbols, and it is
typically used for AGC (automatic gain control), synchronization
and coarse frequency offset estimation. Autocorrelation of the STF
field with a 0.8 .mu.s time shift generates a slowly varying
triangle shaped autocorrelation curve. The triangle's peak provides
an estimate of the frame start time t.sub.STF. However, this
autocorrelation peak is sensitive to noise; moderate noise can move
the autocorrelation peak by 100-200 ns. Furthermore,
autocorrelation with a 0.8 .mu.s time shift can produce false
detection in the presence of tone interference or narrow-band
interference whose period is a sub-multiple of 0.8 .mu.s.
The STF is followed by an LTF, which includes two identical long
training symbols, each 3.2 .mu.s long with a 0.8 .mu.s guard
interval. For received signals that are not subject to fading,
cross-correlation of the long training symbols with the LTF symbol
template generates a sharp peak. The peak provides LTF
synchronization timing and reliability information; the peak
position is not very sensitive to noise while the peak value
generally becomes lower with increasing noise level. LTF
synchronization based on cross-correlation can be used to validate
detections by an STF autocorrelation synchronizer in order to
discard false STF detections, but it is sensitive to the multipath
effect, especially when a late arrival path is strong.
The combination of the above two synchronization methods, coarse
synchronization timing estimation based on STF autocorrelation
followed by verification and fine timing adjustment of LTF
cross-correlation, works well when there is only one transmitter
space time stream.
In order to increase system throughput, according to IEEE
802.11n-2009, multiple transmitter space time streams are
transmitted simultaneously. To prevent unintentional beamforming
when similar signals are transmitted in different space time
streams, two sets of cyclic time shift values are applied to the
non-HT portion and the HT portion of frames respectively. These
cyclic time shift values are defined in the IEEE 802.11n-2009
amendment (see table 20-8 and 20-9). The time shift values can be
as high as -200 ns and -600 ns for non-HT and HT portion
respectively. However, the cyclic time shift introduces a pseudo
multipath problem which can cause cross-correlation based time
synchronisation algorithms to fail. FIG. 1 is an example of failed
LTF synchronization due to the pseudo multipath problem.
The data shown in FIG. 1 is for a receiver operating in HT-GF mode,
with a sample period of 50 ns. Due to the -400 ns cyclic time shift
introduced in the second transmitter space time stream, there are
two peaks occurring in the LTF cross-correlation profile, the late
one being the correct synchronization time. Because the first peak
is stronger than the second, a receiver that uses the strongest
peak to estimate synchronization timing will make a 400 ns
synchronization timing error in this case.
With the increase in the number of transmitter space time streams,
the number of LTF cross-correlation peaks caused by the cyclic time
shift increases, causing their merger or mutual cancellation, so
that the LTF cross-correlation profile becomes more complicated.
This is called the pseudo multipath effect.
Most synchronization methods for OFDM WLAN are either
autocorrelation based algorithms or cross-correlation based
algorithms. For autocorrelation based algorithms, when the SNR is
not very high, the timing error will be large; autocorrelation also
suffers from false detection caused by tone interference or
narrow-band interference whose period is a sub-multiple of 0.8
.mu.s. Conventional cross-correlation based algorithms suffer from
the pseudo multipath effect in MIMO OFDM WLAN.
Other synchronization methods use a maximum likelihood estimation
to achieve better performance; however their complexity is too high
for practical implementation. See E. G. Larsson, et al, "Joint
Symbol Timing and Channel Estimation for OFDM Based WLANs", IEEE
Commun. Letters, vol. 5, no. 8, pp. 325-327, August 2001.
Another class of synchronisation methods relies on the fact that
the signal transmitted during the guard interval of each OFDM
symbol is repeated at the end of the symbol. See T. M. Schmidl and
D. C. Cox, "Robust Frequency and Timing Synchronization for OFDM",
IEEE Trans. On Communications, vol. 45, no. 12, pp. 1613-1621,
December 1997. However, these methods are more suitable for
synchronisation to continuous streams of OFDM symbols, such as in
DAB or DVB, rather than for frame transmissions as defined in IEEE
802.11n-2009.
To tackle the pseudo multipath effect in an 802.11n OFDM WLAN
system, a three-step timing synchronization method is proposed in
Dong Wang, Jinyun Zhang, "Timing Synchronization for MIMO-OFDM WLAN
Systems", IEEE Wireless Communications and Networking Conference,
2007. WCNC 2007, pp 1178-1183. In the first step, a sliding window
differentiator is concatenated with an auto-correlator to remove
the auto-correlation plateau; in the second step, a SIR
(signal-to-interference ratio) metric is calculated based on the
cross-correlation output in a small search window around the
estimated coarse timing position. In the third step, the frame
timing is refined in a small window around the estimation from the
second step. The implementation complexity is high; also the
performance of the algorithm depends heavily on selection of
parameter values.
SUMMARY OF THE INVENTION
According to one aspect, the invention provides a method of WLAN
frame detection in a received signal, wherein the frame comprises
first and second training sequences and the method comprises
auto-correlating the signal with a delayed version of itself to
establish a first frame boundary estimate based on behaviour of the
autocorrelation result due to the inclusion of the first training
sequence in the frame, cross-correlating the signal with a copy of
the second training sequence at a range of time offsets in order to
generate a first cross-correlation profile, classifying the first
cross-correlation profile into one of a number of categories,
establishing a second frame boundary estimate from the first
cross-correlation profile in a manner dependent upon the category
assigned to the first cross-correlation profile and determining a
refined frame boundary estimate on the basis of a consideration of
the first and second frame boundary estimates.
In certain embodiments, the first training sequence is an OFDM WLAN
STF. In certain embodiments, the second training sequence is one
symbol of an OFDM WLAN LTF.
In certain embodiments, the frame comprises a third training
sequence and the method further comprises cross-correlating the
signal with a copy of the third training sequence at a range of
time delays in order to generate a second cross-correlation profile
and boosting the first cross-correlation profile prior to its
classification by coherently adding the second cross-correlation
profile into the first cross-correlation profile. In certain
embodiments, the second and third training sequences are different
symbols of an OFDM WLAN LTF.
In certain embodiments, wherein determining a refined frame
boundary estimate comprises selecting, as the refined frame
boundary estimate, the first frame boundary estimate, the second
frame boundary estimate or a weighted combination of the first and
second frame boundary estimates. The selection of the refined frame
boundary estimate may depend on the peak magnitude in the
auto-correlation result. The selection of the refined frame
boundary estimate may depend on the difference between the first
and second frame boundary estimates.
In certain embodiments, one of the categories is where the first
cross-correlation profile has a single peak and for that category
the second frame boundary estimate is established as the position
of that peak.
In certain embodiments, one of the categories is where the first
cross-correlation profile has just two peaks and for that category
the second frame boundary estimate is established as the position
of the later of those two peaks. The first cross-correlation
profile may be smoothed before establishing the second frame
boundary estimate.
In certain embodiments, one of the categories is where the first
cross-correlation profile has more than two peaks and for that
category the second frame boundary estimate is established in
dependence upon the positions of the three largest peaks.
In certain embodiments, one of the categories is a residual
category for the case where the first cross-correlation profile
fits no other category and for the residual category the second
frame boundary estimate is established as the position of the
maximum in the profile after smoothing.
The invention employs auto-correlation and cross-correlation. These
processes may or may not be normalised, depending on the
requirements of the specific information concerned.
The invention and various embodiments thereof have been described
above in terms of a method. It is to be understood that the
invention extends also to apparatus for, or software (whether
carried by a suitable carrier--such as a memory device--or
otherwise) for, performing a method according to the invention.
BRIEF DESCRIPTION OF THE FIGURES
By way of example only, certain embodiments of the invention will
now be described by reference to the accompanying drawings, in
which:
FIG. 1 is a graph showing failed WLAN OFDM LTF synchronisation due
to cyclic time shift;
FIG. 2, including FIGS. 2a to 2d, is an illustration of examples of
four correlation profiles falling within four different groups;
FIGS. 3a to 3d are illustrations of how a timing error varies with
SNR for different numbers of space time streams in the presence of
an AWGN channel;
FIGS. 4a to 4d are illustrations of how a timing error varies with
SNR for different numbers of space time streams in the presence of
a fading channel; and
FIG. 5 is a schematic block diagram of an OFDM WLAN receiver.
DETAILED DESCRIPTION
FIG. 5 illustrates, schematically, an OFDM WLAN receiver 10. The
receiver 10 comprises an antenna 12, an RF front-end 14, an
analogue-to digital converter (ADC) 16, a data processor 18 and a
memory 20. OFDM signals received at the antenna 12 are filtered,
amplified and down converted in frequency in the RF front-end 14.
The resulting signals are then converted into digital signals by
the ADC 16 that are then supplied to the processor 18. The
processor 18 in conjunction with the memory 20 decodes the
information contained in the OFDM signals. One part of the process
of recovering the information content of the OFDM signals is frame
synchronisation. An algorithm, hereinafter referred to as a
synchronisation algorithm, employed by the processor 18 to detect
OFDM WLAN frame boundaries in digital signals arriving from the ADC
16 will now be described.
Synchronisation Algorithm--Brief Outline
In brief, the synchronisation algorithm comprises the following
five steps.
Step 1: Make a frame detection (STF detection), a first timing
estimate t.sub.STF and a coarse frequency estimate f.sub.STF based
on autocorrelation with an 800 ns time delay. The subsequent
processing steps (2 to 5 described below) are only performed
following an STF detection.
Step 2: Perform the coherent sum of two LTF symbol
cross-correlation profiles, within a time window centred on the
expected timing of the first LTF, estimated based on t.sub.STF. The
first LTF consists of two LTF symbols; the cross-correlation
profiles for these two symbols are coherently added to improve the
reliability of detection. The result of this sum is the LTF
cross-correlation profile used in subsequent processing.
Step 3: LTF cross-correlation profile classification. The LTF
cross-correlation profile is classified into one of four groups.
For each group, the LTF synchronization timing is estimated
differently to reduce the timing error.
Step 4: Perform LTF synchronisation detection and estimate LTF
timing. An FIR filter is proposed to improve timing accuracy for
some groups of LTF cross-correlation profile classification. At
this stage, STF synchronisation detections that are not validated
by LTF synchronisation detections are discarded as false
alarms.
Step 5: Synchronisation timings from STF and LTF synchronizers are
combined to give the final synchronization timing t.sub.sync. When
the received signal is strong (the SNR is high), the STF
synchronizer timing is used, t.sub.sync=t.sub.STF; when the SNR is
low, the LTF synchronizer timing is used, t.sub.sync=t.sub.STF;
when the SNR is neither high nor low, based on the observation that
the STF timing error tends to be late while the LTF timing error
tends to be early,
##EQU00001## Synchronisation Algorithm--a More Detailed
Discussion
Given a discrete-time received base-band signal s(n), for n=1, 2, .
. . , N, where the sample rate is 1/.DELTA.t, the proposed
synchronization algorithm is described in more detail below:
Step 1: Continuously calculate the autocorrelation profile with an
800 ns time delay:
.function..times..function..times..function. ##EQU00002##
When the sample period .DELTA.t is 50 ns, L and T.sub.0 are 144 (9
STF symbols) and 16 samples respectively. Then search for local
peaks of R(n). We denote one such peak position n.sub.0 and its
peak value V.sub.STF. V.sub.STF is compared with a threshold value
T.sub.STF. When V.sub.STF>T.sub.STF, STF synchronization is
detected and the STF synchronization timing estimate is
t.sub.STF=(n.sub.0-L).DELTA.t. If V.sub.STF<T.sub.STF, no STF
synchronization is found and if s(n) contains a WLAN frame starting
at (n.sub.0-L).DELTA.t, then a missed detection occurs. On the
other hand, when s(n) does not contain a WLAN frame, but a peak is
detected with V.sub.STF exceeding T.sub.STF, then a false STF
detection occurs. T.sub.STF is selected empirically to minimise
false detections (`false alarms`) and missed detections.
Step 2: When STF synchronisation is detected (V.sub.STF exceeds
T.sub.STF), then frequency offset compensation is applied to the
received signal prior to further processing as shown in equation
(2). .DELTA.f is the estimated frequency offset; it may be the
coarse frequency offset estimate f.sub.STF or on an adjusted
frequency offset estimate based on processing the LTF field.
s(n)=s(n)exp(-j2.pi.n.DELTA.t.DELTA.f) (2)
Then calculate the LTF cross-correlation profile:
.function..times..function..function..times..function.
##EQU00003##
In the above expression, * is the complex conjugate. The template
{T(k)|k=1, 2, . . . , K} is the ideal LTF symbol defined in the
IEEE 802.11n-2009 amendment or a quantized version of it. When the
sample period is 50 ns, the values of T.sub.1, T.sub.2 and K are
48, 64 and 64 samples respectively. The range for m is [-W W],
where W is the LTF synchronisation timing search radius, for
example 0.8 .mu.s before and after t.sub.STF. The expression in
equation (3) is in fact a coherent sum of the correlation of T with
the first LTF symbol (assumed to commence at symbol
n.sub.0+T.sub.1+m+1 in s) and the correlation of T with the second
LFT symbol (assumed to commence at symbol
n.sub.0+T.sub.1+m+T.sub.2+1 of signal s.
Step 3: Classify the LTF cross-correlation profile into one of the
following four groups: (1) Single peak; (2) Two peaks which are
less than 700 ns away; (3) More than two peaks; (4) The others.
Typical LTF cross-correlation profiles for the above four groups
are illustrated in FIG. 2. Classification is based on parameters
estimated from the LTF cross-correlation profile. Assuming the
values of the three largest LTF cross-correlation profile local
peaks are, in decreasing order of the peak magnitude, P.sub.1,
P.sub.2, P.sub.3, and their time positions are d.sub.1.DELTA.t,
d.sub.2.DELTA.t, d.sub.3.DELTA.t respectively
(-W.ltoreq.d.sub.k.DELTA.t.ltoreq.W, k=1, 2, 3) and with a mean
value m.sub.0 of the cross-correlation profile, then the parameters
used for classification are:
.times..DELTA..times..times. ##EQU00004##
The classification criteria are described below: (1) If
<.ltoreq..times..times..times..times..gtoreq. ##EQU00005## the
profile belongs to the single peak group. For example, in our
simulations, T.sub.1=2.5, r.sub.1=0.35, T.sub.2=4; (2) Otherwise,
if
< ##EQU00006## and P.sub.31.ltoreq.r.sub.3 and either
.gtoreq. ##EQU00007## or P.sub.21.gtoreq.r.sub.2 and
d.sub.21.ltoreq.700 ns, the profile belongs to the two peak group.
In the simulation, T.sub.3=2.25, r.sub.3=0.5, r.sub.2=0.4; (3)
Otherwise if P.sub.31.gtoreq.r.sub.3, it belongs to group (3); (4)
Otherwise, the profile belongs to group (4).
Step 4: After classification, the LTF synchronization timing
t.sub.LTF is determined as below:
(1) If a synchronization is found by the STF synchronisation in
step 1 but P.sub.1<P.sub.th, then it is considered that false
detection has occurred, which may be caused by tone interference or
narrow-band interference whose period is a sub-multiple of 0.8
.mu.s. The threshold value P.sub.th is chosen to balance the
probabilities of missed detection and false detection.
(2) Otherwise if the cross-correlation profile belongs to group
(1), the synchronisation timing corresponds to the largest
cross-correlation profile peak position,
t.sub.LTF=t.sub.STF+d.sub.1.DELTA.t;
(3) Otherwise if the cross-correlation profile belongs to group
(2), the cross-correlation profile is smoothed by an FIR filter
whose impulse response is [1 0.5 0.5 0.5 0.5]. This filter enhances
late peaks in the L-LTF when there are multiple space time streams.
Then denoting d.sub.2' the position of the second local peak in
terms of time sequence of the smoothed profile,
t.sub.LTF=t.sub.STF+d.sub.2.DELTA.t.
(4) Otherwise if the cross-correlation profile belongs to group
(3), assuming that, among the three largest local peaks, the
earliest peak position is d.sub.pe and the latest peak position
d.sub.pl, the LTF synchronization timing t.sub.LTF is:
.times..DELTA..times..DELTA..times..times..times..times..times..DELTA..ti-
mes..times..DELTA..ltoreq..times..times..times..DELTA..times..times..times-
..times..times..times.<.times..DELTA..times..times..DELTA..ltoreq..time-
s..times..times..DELTA..times..times. ##EQU00008##
(5) Otherwise if the cross-correlation profile belongs to group
(4), pass the LTF cross-correlation profile through the FIR filter
mentioned in (3), find the largest local peak position {circumflex
over (d)}.sub.1.DELTA.t in the smoothed profile and use
t.sub.LTF=t.sub.STF+{circumflex over (d)}.sub.1.DELTA.t.
Step 5: The STF and LTF timings are used to determine the final
synchronization timing:
(1) If the STF autocorrelation profile peak V.sub.STF is high, the
LTF cross correlation profile peak is also very high and the
difference between t.sub.STF and t.sub.LTF is no more than 200 ns,
then the LTF synchronization timing t.sub.LTF is adopted as the
final synchronisation timing;
(2) Otherwise if the STF autocorrelation profile peak is high, and
either the LTF cross-correlation profile peak is not very high, or
the difference between t.sub.LTF and t.sub.STF is more than 200 ns,
then STF synchronization timing t.sub.STF is adopted as the final
synchronisation timing;
(3) Otherwise if the STF autocorrelation profile peak is neither
high nor low, and the difference between t.sub.LTF and t.sub.STF is
no more than 200 ns, the LTF synchronizer timing t.sub.LTF is used
as the final synchronisation timing;
(4) Otherwise if the STF autocorrelation profile peak is neither
high nor low, but the difference between t.sub.LTF and t.sub.STF is
more than 200 ns, then
##EQU00009## is used as the final synchronisation timing;
(5) Otherwise if the STF autocorrelation profile peak is very low
and the difference between t.sub.LTF and t.sub.STF is more than 200
ns, then the LTF synchronizer timing t.sub.LTF is used as the final
synchronisation timing.
Step 5 depends on, amongst other things, tests using relative
criteria, e.g. examining whether peaks in correlation process
results are high or low. The meanings of the terms "high" and "low"
in the context of step 5 are determined empirically (and in any
event depend on whether the auto- and cross-correlations are
normalised (in the preceding embodiment they are not)). For
example, it is possible to estimate the probability distribution of
the peaks in the auto- and cross-correlation profiles (e.g. through
Monte-Carlo simulation over a range of relevant SNRs and fading
conditions) and define thresholds for determining whether a peak is
"high" or "low".
The variations in the synchronization ratio (sync ratio) with SNR
for various numbers of transmitter space time streams under an AWGN
channel and a fading channel (the IEEE fading model C) are shown in
FIGS. 3a to 3d and FIGS. 4a to 4d respectively. In these figures,
HT-MM means mixed mode high throughput mode, HT-GF means Greenfield
high throughput mode; TxX means X transmit antennas; STBC0 or STBC1
means disable/enable spatial time block coding; Q1 means using
identity matrix of direct mapping in spatial mapping (this
definition refers to paragraph 20.3.11.10.1 `Spatial mapping` of
the IEEE 802.11n-2009 amendment). STS is an abbreviation for `space
time stream`. In these simulations, the sample period is 50 ns. The
legends in FIGS. 3 and 4 represent tolerance windows for the final
synchronisation timing error. So [-n.sub.1 n.sub.2] means the
synchronization timing error is
-n.sub.1.DELTA.t.ltoreq..DELTA.t.sub.sync.ltoreq.n.sub.2.DELTA.t.
The `sync ratio` in these figures is the percentage of simulation
runs whose synchronization timing errors fall within the prescribed
range.
As shown in the above FIGS. 3a to 3d, when the SNR is high, under
an AWGN channel, the timing error will be zero or very close to
zero. Over the entire SNR range (down to an SNR of 2 dB), under an
AWGN channel, the timing error is within [-2 2] samples ([100 ns
100 ns]).
The proposed synchronization algorithm also works well under
moderate fading conditions (IEEE fading model C) as shown in FIGS.
4a to 4d.
Thus, a frame synchronization algorithm for MIMO OFDM WLAN has been
described, which algorithm can effectively mitigate the pseudo
multipath effect caused by cyclic time shift under both an AWGN
channel and a fading channel.
The described frame synchronisation algorithm has low complexity
for real time implementation; it effectively overcomes the pseudo
multipath effect due to the cyclic time shift in 802.11n MIMO OFDM
WLAN. Through simulation, the synchronization algorithm has shown
small timing error, so that receipt of frames using short guard
interval mode can work properly.
* * * * *